Measuring online advertising transparency in singapore

Measuring Online Advertising Transparency in Singapore: An Investigation of Threats to Users Dr Paul A Watters, University of Ballarat OCTOBER 2013 Electronic copy copy available available at: Electronic at:https://ssrn.com/abstract=2362626 http://ssrn.com/abstract=2362626 CONTENTS EXECUTIVE SUMMARY DEFINITIONS 5 INTRODUCTION 6 METHODOLOGY 10 RESULTS 12 CONCLUSION 20 REFERENCES 22 APPENDIX A 24 APPENDIX B 38 APPENDIX C 46 APPENDIX D 48 measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy copy available available at: Electronic at:https://ssrn.com/abstract=2362626 http://ssrn.com/abstract=2362626 EXECUTIVE SUMMARY Recent studies have highlighted the role that internet advertising plays in supporting the revenue of rogue websites (Taplin, 2013) Certainly, such advertising generates enormous profit margins for operators of these websites, and present an ongoing threat to the viability of Singapore’s creative industries However, a recent study by Watters (2013) investigated the harms to users from viewing the increasingly “high risk” nature of advertising being hosted on these sites An analysis of advertising transparency was undertaken using the methodology developed by Watters (2013) A total of 5,000 webpages representing ten pages sampled from Google’s ad transparency report were downloaded in Singapore, and each ad banner categorised as being High Risk or Mainstream, where each page was verified as being in breach of Digital Millennium Copyright Act (DMCA) for movies and TV shows 10% of ads were Mainstream, 90% were High Risk The prevalence of Mainstream ads being served to Singaporeans is one order of magnitude greater than similar advertising being shown to Australians The policy implications of this result and future research directions, including methodology enhancements, are discussed KEY FINDINGS: • For the Singaporean population, only 10% of advertising on rogue websites was for mainstream businesses 90% of the advertisements displayed on rogue websites were categorised as High Risk (i.e Malware, sex industry, gambling, scams and downloading sites) • With 44% of advertisements on rogue websites being categorised as malware, Singaporeans who access rogue websites are at a substantially higher risk of being exposed to malware infection • 10% of advertisements on rogue websites were categorised as gambling These gambling advertisements, targeting Singaporean users, operate outside Singapore’s jurisdiction and regulations DEFINITIONS INTERNET ADVERTISER A business, government, association or individual that desires to sell goods or services, or provide information to, a target group of consumers Internet advertising competes with traditional advertising for marketing budgets Singapore’s online advertising market was valued at US$107m in 2011 and is growing rapidly1 http://www.asiamediajournal.com/pressrelease.php?id=3906 Electronic copy available at: https://ssrn.com/abstract=2362626 INTERNET ADVERTISING Ads are typically placed as “banners” on a website, which direct a user to another site when clicked The contents of the ad are similar to a highway billboard, except that the can incorporate interactive elements such as animation Ads on the same page are often rotated through a predetermined or random sequence, depending on the advertising plan that an advertiser has subscribed to While some sites host and manage their own banners, most often, these are managed by a thirdparty advertising network These ad networks act as an intermediary between an advertiser and many hundreds, thousands or millions of sites, allowing an advertiser to increase their reach to potential consumers while only dealing with a single agency Advertisers typically operate either a “pay per impression” or “pay per click” model, billing an advertiser every time a user views or clicks on a banner ad respectively MAINSTREAM ADVERTISING Mainstream ads are those placed by legitimate businesses that operate within the formal economy Such businesses operate through a corporate structure and offer goods or services which fall outside the black market, grey market or underground economy HIGH-RISK ADVERTISING High-Risk ads are those promoting goods or services which fall outside the legitimate economy or white market, may be illegal or restricted within certain jurisdictions but not others, or may be fake or counterfeit Examples include the sex industry, gambling and suspicious software/malware, such as antivirus software which actually installs a Trojan Horse on a user’s system Many of the ads are likely to fall into scam categories described by Stabek et al (2009) ADVERTISING NETWORK Ad networks facilitate the placement of an advertiser’s ads on numerous websites according to a specific revenue model Ad networks specialise in anticipating consumer’s needs and wants by building up profiles of users who click most frequently on certain ad categories on certain page themes, which can lead to more targeted, personalised, and relevant advertising For the purposes of this paper, sites that host advertising on behalf of external / third-party advertisers are also grouped under this category, even if they only provide banners on sites within their own domain For example, isohunt.com provides their own ad network exclusively for their own site, and not to other sites; they also host banners from other ad networks ROGUE SITE A website which provides an index and search capability for torrents of infringing content, a “file locker” site which provides hosting for such material, or a “link site” which provides direct links to content on third party sites The primary motivation for users visiting these websites is to access infringing content These sites can all use advertising as either primary or secondary sources of income DIGITAL MILLENNIUM COPYRIGHT ACT (DMCA) The DMCA provides ISPs with indemnity against liability for copyright infringement, provided that they agree to cooperate in “takedowns” of material which is alleged to be infringing, typically after being notified by a rightsholders Google provides a report of requests that they have received and actioned on behalf of rightsholders in order to provide transparency to their users measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 INTRODUCTION Online advertising has a 20 year long history (Medoff, 2000), progressing from simple ad banners displayed on a fixed rotation schedule, through to personalised, behavioural advertising networks, which use profiles of individual users to present the most “relevant” advertisements (McStay, 2011) Such technologies make extensive use of “tracking cookies” (Watters, 2012) and the linkages between advertising networks and cookies have recently been monitored and explored for the most popular websites in Singapore (Herps et al, submitted) The most interesting result from this study was that the number of cookies stored on a user’s computer from any of the Top 50 most-visited sites for Singaporeans ranged between and 86 The sophistication and the extent to which user behaviour is tracked and experiences customised is only going to increase over time, as is the overall volume of advertising Indeed, in 2012, online advertising spending in the US reached US$39.6b, exceeding the amount spent on traditional print advertising for the first time (eMarketer, 2012) Furthermore, some companies are in a unique position to know “everything” about their customers Google, for example, has the capacity to monitor almost all of the world’s information, including personal emails, YouTube movies, Android phones, news services, images, shopping, blogs and so on (Cleland, 2013) Through its acquisition of Doubleclick, Google controlled an estimated 69% of the online advertising market (Browser Media, 2008), however, the rise of social media advertising (especially through Facebook) has seen this reduce to 56% (Womack, 2013) Clearly, there is a potential confluence of capability and opportunity to maximise the number of “eyeballs” exposed to online ads What are the implications of this massive rise in advertising expenditure, which coincides with an increased ability for online advertising networks to be able to best “place” ads to suit specific customers? One particular type of website – those associated with file sharing of infringing content – appears to have wholeheartedly embraced advertising Indeed, advertising revenues provide the commercial motivation for criminal syndicates to operate such ‘rogue’ web sites While the connection between film and television piracy and organised crime has been explored elsewhere, in terms of direct revenues (Treverton et al, 2009), there has been far less publicity about the advertising revenues generated from sites that appear to offer infringing content for free, or at least, offer torrents that enable users to download such material Certainly, the links between the underground economy and the internet have been criticised for facilitating sexual exploitation and human trafficking through organised crime – in the classic paper in this field, Hughes (2000) highlighted how global advertising and marketing of prostitution have led to increases in volume globally Furthermore, Hughes identified that a lack of regulation of internet advertising was the key policy failure in preventing harm to women and children The Pirate Bay is one of the most popular sites for providing torrents to infringing content, and has been the subject of criminal proceedings against its operators in Sweden In the 2009 trial of its operators, their expenses were estimated to be US$110,000 p.a (Olsson, 2006; Kuprianko, 2009), with advertising revenues in the order of US$1.4m p.a Electronic copy available at: https://ssrn.com/abstract=2362626 (Sundberg, 2009) – in other words, an extremely profitable business with gross margins of 1272%! A recent study (Detica, 2012) indicated that there are six different business models operating within the pirate site marketplace, ranging from advertisement and donation funding, through to subscriptions and freemium sites, where subscribers can gain faster access to illicit content by paying a subscription fee 83% of the sites in that study operated using a central website Selling advertising on file locker and torrent search sites is the major source of revenue for such sites The Pirate Bay, for example, regularly features in the Top 50 sites accessed by Singaporeans (as computed by alexa.com) , and so it is a potentially attractive space for advertisers and ad networks, since the number of potential “eyeballs” is very high Maximising “eyeballs” leads to clicking, which drives revenue for the ad networks (if they operate a Pay Per Click revenue model), and sales for the advertisers A key question for advertisers and ad networks is the extent to which they wish to be associated with this type of activity; indeed, due to the complex algorithms which decide which ads to display to which users, advertisers may not be aware of every site that their ads are being displayed on Being able to quantify the scale of advertising on these sites is important, since informing and making advertisers aware of the integrity of the sites on which their ads are being displayed can then be undertaken Advertisers will thus be able to make more informed choices about their use of online advertising networks (the companies who provide aggregation of space on web sites) who are supporting piracy by selling ad space on torrent and file locker sites A recent set of best practice guidelines for ad networks to address piracy and counterfeiting have recently been released2, and early indications are that most of the world’s major web companies will participate3 There have been few systematic studies investigating the relationship between piracy and advertising, and most have been concerned with the impact of interventions to reduce piracy For example, Sheehan et al (submitted) identified that increasing the perception of legal risk for college students was most likely to influence downloading behaviour, while Gopal et al (2009) weighed up the ethical predispositions of downloaders and their beliefs in justice and law to the money potentially saved by downloading infringing content Indeed, it is this appeal to justice as the primary virtue of social behaviour (Rawls, 1999) that may concern ethical advertisers if their advertising expenditure was being used to fund illicit activities Recently, the USC Annenberg Lab has begun producing a report that explores the relationship between piracy sites and online advertising networks (Taplin, 2013) The USC report provides a method for revealing the advertisers whose ads are most likely to be served up on these sites, which may be occurring without the direct knowledge of the advertiser While the objectives of USC research are significant, the monthly rankings of the “top ten” advertising networks responsible for placing the most ads on web sites that support infringing content are surprisingly variable – Google, for example, was ranked at #2 in January 2013, but did not appear at all in 2 http://2013ippractices.com/bestpracticesguidelinesforadnetworkstoaddresspiracyandcounterfeiting.html 3 http://torrentfreak.com/tech-giants-sign-deal-to-ban-advertising-on-pirate-websites-130715/ measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 the February and March 2013 lists at all One interpretation of the result could be that the January report achieved its goal of sensitising advertising networks, and that Google subsequently withdrew from placing ads on those sites Alternatively, the variation could be due to biases inherent in studies using an observational methodology, including: • Selection bias, in the way that infringing sites are selected The study uses a single source (the Google Transparency Report of domains with the most DMCA takedown requests), rather than using a consensus technique which combines the ranks of several different data sources to provide the most accurate ranking This type of triangulation is commonly used in observational studies as a form of triangulation; • Information bias, since only one technique for collecting data is used (HTML and JavaScript code scraping), where other techniques may be more accurate or representative of advertising behaviour For example, persistent cookies have been strongly associated with behavioural advertising, and the frequency of tracking cookies being stored by ad networks could provide an alternative measure of presence of significance Yet the USC report does not analyse cookies at all; and • Recall bias, since the data analysed was only from English- language websites and advertising networks which may potentially have a higher level of visibility than networks which operate in other geographic zones, languages, encoding types etc Also, the lack of detail in how measures like the “top 500” sites prevent the study results from being directly replicated, which would be the standard required for peer review by other researchers By not providing this level of detail, the credibility of the USC report may be called into question by the very vocal critics of any research in the anti-piracy field In this paper, we present a more rigorous and fully replicable methodology which should provide a much clearer view of advertising network behaviour in different countries, jurisdictions, languages etc In this study, we specifically target English language films and television shows viewed by Singaporean users; the methodology itself is sufficiently general that it could be applied to any country and any category, including music, computer games, e-books etc A previous study using this methodology focused on the ads being served to Australians In that study, it was found that 99% of the ads from the “top 500” sites were High-Risk, while only 1% were Mainstream It is predicted that a similar composition will be found in Singapore Electronic copy available at: https://ssrn.com/abstract=2362626 Electronic copy available at: https://ssrn.com/abstract=2362626 september, 2013 METHODOLOGY The main goal of the methodology is to identify the advertising networks and advertisers from a sample of DMCA complaints, which have been ranked in terms of the number of complaints upheld by Google (through their Transparency Report) These complaints typically relate to the availability of search results for a wide range of potentially infringing content; by only selecting the most complained about and subsequently upheld complaints as assessed by a third-party (Google), the results should be robust against criticisms that there is no proof that the sites in question were hosting torrents of infringing content or infringing content directly, in the case of a file locker site The methodology operates by downloading each page from the “top 500” complaints submitted to Google within the previous month, ordered by the number of upheld complaints Since each DMCA notice can contain many thousands of individual URLs, a sampling procedure can be used to identify a representative subset of URLs, and the advertisements on each page can be downloaded along with their metadata In the case of simple banner ads, it is then relatively easy to identify the advertisers concerned; in the case of each distinct advertisement, a rule can be generated using SQL or similar to identify all advertisements with the same metadata However, some advertising networks use JavaScript obfuscation and a series of redirects to obscure the ultimate destination for the advertising banner; in this case, manual inspection must be performed, in the absence of a general purpose image/logo recognition system The overall prevalence of a particular advertiser on each network can be then be computed and ordered by frequency A key question arises concerning the relevance of using DMCA notices from Google as the basis for determining prevalence in Singapore Singapore’s population has a linguistically diverse Chinese (75%), Malay (14%) and Indian (8.8%) population, and given that DMCA notices predominantly relate to English-language movies and television, are these notices relevant? However, in practice – as noted by Ang (2007), English-language television and other forms of media are still dominant Furthermore, it may be of interest to separate out “Mainstream” advertisements as opposed to “High-Risk” advertising, since the Annenberg reports indicate a flight by Mainstream advertising this year from sites that host infringing content Advertisers who may otherwise be unable to place their ads on a Mainstream site can then take advantage of increasing “eyeballs” by occupying display space Results are thus reported for the High-Risk and Mainstream categories, with the former including categories such as: • Sex Industry, which includes adverts for: »» Penis length extension medication »» Fake personal/dating sites 10 Electronic copy available at: https://ssrn.com/abstract=2362626 »» Pornography of various kinds »» Dating and “foreign bride” sites • Online Gambling • Malware, including »» Fake software incorporating Trojan horse malware (numerous alerts were raised by anti-virus software during the data collection process due to “drive by downloads” of malware) »» Fake anti-virus or anti-scamware »» Suspicious software such as fake video codecs or video players that replicate existing functions within Microsoft Windows The purpose of such downloads is unclear, although it is possible that they could host Trojans or provide backdoor access to systems • Scams, as defined by Stabek et al (2010), such as: »» Premium rate SMS scams »» Fake competitions where no prizes are offered »» Investment scams »» Employment scams The algorithm works as follows: A data collection system is installed physically or logically to attract advertising for a specific geographical/country segment For this study, Singapore was selected The current Google Transparency Report4 is downloaded, which lists all of the DMCA requests for a specific time period5 This list provides one means of identifying sites involved in sharing pirated material The dataset is sorted by the number of URLs removed, retaining the “top 500” DMCA requests (the request list) by complaint category For this study, the complaint category was movies and TV shows; other complaint categories such as pirated software, adult material, music etc were excluded For each report in the request list first 10 URLs are extracted as a representative sample of all of the URLs contained within the report This gives a total of 5,000 web pages to be downloaded (the sample) https://www.google.com/transparencyreport/removals/copyright/data/ The DMCA list for May-July 2013 was used in this analysis measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 http://www.chillingeffects.org/notice.cgi?sID=996407 http://www.chillingeffects.org/notice.cgi?sID=996412 http://www.chillingeffects.org/notice.cgi?sID=996420 http://www.chillingeffects.org/notice.cgi?sID=996912 http://www.chillingeffects.org/notice.cgi?sID=996915 http://www.chillingeffects.org/notice.cgi?sID=997769 http://www.chillingeffects.org/notice.cgi?sID=997801 http://www.chillingeffects.org/notice.cgi?sID=997802 http://www.chillingeffects.org/notice.cgi?sID=997803 http://www.chillingeffects.org/notice.cgi?sID=997805 http://www.chillingeffects.org/notice.cgi?sID=999830 http://www.chillingeffects.org/notice.cgi?sID=999832 http://www.chillingeffects.org/notice.cgi?sID=999898 40 NBCUniversal NBCUniversal NBCUniversal Disney Enterprises Inc Disney Enterprises Inc Fox Fox Fox Fox Fox Fox Fox Fox Electronic copy available at: https://ssrn.com/abstract=2362626 Electronic copy available at: https://ssrn.com/abstract=2362626 september, 2013 APPENDIX B- FULL LIST OF ADVERTISING NETWORKS DETECTED Advertising Network Frequency propellerads.com pobieramy24.pl fhserve.com filestube.com isohunt.com sumotorrent.com adcash.com propellerpops.com btarena.org velmedia.net admxr.com adshost1.com googletagservices.com adshost2.com tlvmedia.com torrentfunk.com adexprt.com pl.bbelements.com google.com webmasterbond.com facebook.com zedo.com popads.net ad4game.com netsprint.eu usualmedia.com uasdel.com coolmirage.com bidvertiser.com torrentroom.com esoft.me fulldls.com sceper.ws lulzimg.com yimg.com cpm24.pl 42 1647 718 702 392 322 279 268 246 240 232 228 203 200 200 184 164 161 138 138 120 110 106 96 93 92 72 72 70 68 64 64 62 60 60 51 46 Electronic copy available at: https://ssrn.com/abstract=2362626 cpmleader.com impresionesweb.com ad6media.fr aclantis.com adserve.com adbooth.net torrentportal.com 1phads.com zoink.it ad4mmo.pl reduxmediia.com isohits.com torrage.com depositfiles.com torrentreactor.net cdnfile.com limetorrents.com clicksor.com onclickads.net lumovies.com cdn.turbobit.net ffdownloader.com advertise.com adlure.net adsbyisocket.com bitsnoop.com torrentdownloads.me fileserve.com adbrite.com dreamboxcart.com mechodownload.com adtransfer.net fenopy.com torrents.net rapidgator.net downtwarez.com juicyads.com torlock.com torrenthound.com fenopy.se smowtion.com 46 45 42 42 40 37 37 34 34 33 33 32 32 30 30 28 27 27 26 24 24 23 23 23 22 21 21 20 20 20 20 19 19 18 18 17 17 16 16 16 15 measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 monova.org 1337x.org sanatas.com kickasstorrents.com madadsmedia.com campus-party.org h33t.com yourbittorrent.com uptobox.com extratorrent.com extabit.com evolutionmedia.bbelements.com yieldmanager.com friendlyduck.com seedpeer.me newtorrents.info torrentday.com katmirror.com warez-home.net procontentservice.com pubdirecte.com lzjl.com yesads.com w3.org auto-ping.com engine.4dsply.com vidics.ch go.goldbachpoland.bbelements.com torrentzap.com SceneTime.com mrgreen.com predictad.com tinypic.com baypops.com rarbg.com darkmachine.pl ioptionpartners.com zimabdk.com shabakti.com simplyserve.me criteo.com 44 15 15 15 15 15 15 14 14 14 14 14 14 13 13 13 13 13 12 12 12 12 12 12 12 12 12 11 11 11 11 11 11 11 11 10 10 10 10 10 10 Electronic copy available at: https://ssrn.com/abstract=2362626 schema.org onvertise.com v2cigs.com torrentbit.net torrentdownloads.net bubblesmedia.ru tblamnetwork.com serveads.info tvrage.com wikipedia.org torrentcrazy.com xtendmedia.com btjunkie.org thepiratebay.sx katproxy.com pirateproxy.net engine.trklnks.com networkhm.com dt00.net yllix.com btscene.org vatgia.com vidxden.com feedburner.com movie25.com cpxinteractive.com adsopx.com ahashare.com digg.com kovla.com icio.us extra33.com bittorrent.am imdb.com twitter.com mypcbackup.com phaze.co torrent.cd fenopy.eu c8.net.ua stumbleupon.com 9 9 8 8 8 8 7 7 7 6 6 6 6 5 5 5 4 4 4 Electronic copy available at: https://ssrn.com/abstract=2362626 september, 2013 medleyads.com katzddl.ws gssp-a.com quaivatdienanh.com iwannawatch.ch zap2it.com thefutoncritic.com desijuke.net arbopl.bbelements.com opensubtitles.org rotator.hadj7.adjuggler.net bt-chat.com cpxcenter.com btmon.com wordpress.com alivetorrents.com ipodnova.tv ad-center.com exoclick.com gmpg.org moviemotion.info hagalepues.net na5.netdna-cdn.com static.flickr.com thepiratebay.org refban.com adhood.com seedpeer.com raidrush.org yesads.com, dirtywarez.com redown.se kat.ph torrentco.com linksbox.net smashingmagazine.com eztv.it vertor.com btguard.com rmbn.net 46 4 4 4 4 4 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 Electronic copy available at: https://ssrn.com/abstract=2362626 us.e-planning.net www.done movie-stars.us adnxs.com altervista.org torrentbar.com cs-puchatek.pl abc.go.com googleapis.com gumgum.com twilight.ws netz.ru abcfamily.go.com torrentlocomotive.com imagetoupload.com best-top.biz uploadbaz.com adpv.com blamcity.com downloadstube.org awempire.com RARBG.com.txt esoft.ws Torrentday.com.txt addthis.com am11.ru Speed.Cd 205.157.43 rlslog.net Speed.Cd.txt postimage.org popcash.net movietorrents.eu played.t fastpic.ru mybittorrent.com ro2.biz vemba.com mightyupload.com iwannawatch.net mininova.org 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 bitshare.com www.uplo scenetime.com.txt lumfile.com www.zala ilibr.org:6969 technorati.com doubleclick.net go.com prq.to nuseek.com Cpasbien.me] 48 1 1 1 1 1 1 Electronic copy available at: https://ssrn.com/abstract=2362626 Electronic copy available at: https://ssrn.com/abstract=2362626 september, 2013 APPENDIX C- HIGH-RISK ADVERTISING EXAMPLES SCAMS, GAMBLING & MALWARE 50 Electronic copy available at: https://ssrn.com/abstract=2362626 SEX INDUSTRY measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 APPENDIX D - MAINSTREAM ADVERTISING EXAMPLES FINANCE/BANKING FOOD/CONSUMER 52 Electronic copy available at: https://ssrn.com/abstract=2362626 TOURISM measuring online advertising transparency in singapore: an investigation of threats to users, dr paul watters, icsl, university of ballarat Electronic copy available at: https://ssrn.com/abstract=2362626 ACKNOWLEDGEMENTS The ICSL receives funding from Westpac, IBM, the Australian Federal Police, the Attorney General’s Department and the Telematics Trust Electronic copy available at: https://ssrn.com/abstract=2362626 ... Lead in Online- Ad Market Downloaded from http://www.bloomberg.com/news/2013-06-13/ google-is-projected-to-expand-lead -in -online- ad-market.html measuring online advertising transparency in singapore: ... HIGH-RISK ADVERTISING EXAMPLES SCAMS, GAMBLING & MALWARE 50 Electronic copy available at: https://ssrn.com/abstract=2362626 SEX INDUSTRY measuring online advertising transparency in singapore: an investigation... doubleclick-deal-means-google-controls-69-of-the -online- ad-market/ Cameron, N (2013) Singapore? ??s online advertising market valued at $17.1bn Downloaded from http://www.cmo.com.au/article/466022/ Singapore _online_ advertising_ market_valued_17_1bn/